The quick read: • In an amended IPO filing on Tuesday, Nov. 10, 2026, Anthropic said it "made a determination that we have met this threshold": the line in its Responsible Scaling Policy for AI accelerating its own research. • Per Yahoo Finance, the threshold is AI doing enough of Anthropic's research to roughly double the pace of progress: about two years of progress, at the 2018 to 2024 rate, packed into one. • Anthropic says the call rests on what it expects from future models and its overall rate of progress, "rather than on what we have observed from current models." It says its use of AI "has already been accelerating our research and development." • It says it does not expect to safely use Claude to fully autonomously build the next generations of frontier models without meaningful human oversight, and warns it "may lose visibility" across some processes as it leans on Claude. • In August, Anthropic argued it was still below the threshold. On Nov. 5, the outside reviewer METR called that argument "overstated" and said a crossing by mid-July was "plausible," per Yahoo Finance. • Anthropic's side: crossing the line means more safeguards, not a shutdown, and it will release further reports. The IPO is set to price Wednesday, Nov. 18, at $70 to $78 a share, up to about $1.4 trillion fully diluted, and trade Nov. 19 on Nasdaq as ANT.
“Anthropic Says Its AI Is Speeding Up How It Builds AI, Days Before Its IPO”
who’s making the noise
Anthropic's amended S-1 of Nov. 10, 2026, as reported by Yahoo Finance (Nov. 10 and Nov. 12) and CASRAI (Nov. 12).
our take
The company selling its stock next week is telling buyers, in the same document, that its AI now speeds up the building of its next AI fast enough to cross a safety line it drew for itself. That is a disclosure, not an accident: Anthropic set the line, reported crossing it, and says its response is more safeguards. But the outside reviewer it relies on says Anthropic's earlier claim that it was still under the line was overstated, and the filing itself warns the company may lose sight of parts of its own work as Claude does more of it. The fair point for Anthropic: no other lab files this kind of warning in public, the determination is based on expected models rather than a specific incident, and it says it will keep humans in control of building new models. The plain move for a small business: nothing in your Claude account changes this week. If you run work through Claude, read the model and data terms you already agreed to, and keep a second provider tested for the jobs that matter. Not investment advice.